causal modeling of fmri data
DESCRIPTION
Causal Modeling of FMRI Data. Joe Ramsey With enormous help from Clark Glymour as well as Russell Poldrack, Stephen and Catherine Hanson, Stephen Smith, Erich Kummerfeld , Ruben Sanchez, and others. Goals. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/1.jpg)
1
Causal Modeling of FMRI Data
Joe RamseyWith enormous help from Clark Glymour as well as
Russell Poldrack, Stephen and Catherine Hanson, Stephen Smith, Erich Kummerfeld, Ruben Sanchez, and others.
![Page 2: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/2.jpg)
2
Goals
1. From Imaging data, to extract as much information as we can, as accurately as we can, about which brain regions influence which others in the course of psychological tasks.
2. To generalize over tasks
3. To specialize over groups of people.
![Page 3: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/3.jpg)
3
What Are the Brain Variables?
In current studies, from20,000 + ………………………………….3
voxels ROIs
ROI = Region of interest
Question: How sensitive are causal inferences to brain variable selection?
![Page 4: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/4.jpg)
4
How are ROIs constructed (FSL)?
• Define an experimental variable (box function).• Use a generalized linear model to determine which
voxels “light up” in correlation with the experimental variable.
• Add a group level step if voxels lighting up for the group is desired.
• Cluster the resulting voxels into connected clusters.– Small clusters are eliminated.– Remaining clusters become the ROIs.– Symmetry constraints may be imposed.
![Page 5: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/5.jpg)
5
Search Complexity: How Big is the Set of Possible Explanations?
X YX YX YX YX Y
}For N variables:
8
2N
For graphical models:
![Page 6: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/6.jpg)
6
Statistical Complexity• Graphical models are untestable unless
parameterized into statistical models.• Incomplete models of associations are likely to
fail tests.• Multiple testing problems.• Multiple subjects/Missing ROIs.• No fast scoring method for mixed ancestral
graphs that model feedback and latent common causes.
• Weak time lag information.
![Page 7: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/7.jpg)
7
Measurement Complexity• Sampling rate is slower than causal interaction
speed.• Indirect measurement creates spurious
associations of measured variables:N1 N2 N3 X1 X3X1 X2 X3 X2
Neural N, measured X Regression of X3 on X1, X2
![Page 8: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/8.jpg)
8
Specification Strategies
1. Guess a model and test it.2. Search the model space or some restriction
of it.a. Search for the full parameterized structureb. Search for graphical structure alonec. Search for graphical features (e.g, adjacencies)
![Page 9: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/9.jpg)
9
What Evidence of What Works, and Not?
• Theory.– Limiting correctness of algorithms (PC, FCI, GES,
LiNGAM, etc., under usually incorrect assumptions for fMRI).
• Prior Knowledge– Do automated search results conform with
established relationships?• Animal Experiments (Limited)• Simulation Studies
![Page 10: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/10.jpg)
10
Brief Review: Smith’s Simulation Study
• 5 to 50 variables• 28 simulation conditions, 50 subjects/condition.• 38 search methods• Search 1 subject at a time.
![Page 11: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/11.jpg)
11
Methods tested by Smith• DCM, SEM excluded; no search. (Not completely true.)• Full correlation in various frequency bands• Partial correlation• Lasso (ICOV)• Mutual Information, Partial MI• Granger Causality• Coherence• Generalized Synchronization• Patel’s Conditional Dependence Measures
– P(x|y) vs P(y|x)• Bayes Net Methods
– CCD, CPC, FCI, PC, GES• LiNGAM
![Page 12: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/12.jpg)
12
Smith’s Results
• Adjacencies:– Partial Correlation methods (GLASSO) and several
“Bayes Net” methods from CMU get ~ 90% correct in most simulations.
• Edge Directions– Smith: “None of the methods is very accurate, with
Patel's τ performing best at estimating directionality, reaching nearly 65% d-accuracy, all other methods being close to chance.” (p. 883)
– Most of the adjacencies for Patel’s τ are false.
![Page 13: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/13.jpg)
13
Simulation conditions (see handout)…
![Page 14: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/14.jpg)
14
SIMULATION 2 (10 variables, 11 edges)
![Page 15: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/15.jpg)
15
Simulation 4 (50 variables 61 edges)
![Page 16: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/16.jpg)
16
Simulation 7: 250 minutes, 5 variables
![Page 17: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/17.jpg)
17
Simulation 8: Shared Inputs
![Page 18: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/18.jpg)
18
Simulation 14: 5-Cycle
![Page 19: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/19.jpg)
19
Simulation 15: Stronger Connections
![Page 20: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/20.jpg)
20
Simulation 16: More Connections
![Page 21: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/21.jpg)
21
Simulation 22: Nonstationary Connection Strengths
![Page 22: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/22.jpg)
22
Simulation 24: One Strong External Input
![Page 23: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/23.jpg)
23
![Page 24: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/24.jpg)
24
![Page 25: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/25.jpg)
25
Take Away Conclusion?
• Nothing works!• Methods that get adjacencies (90%) cannot get
directions of influence.• Methods that get directions (60% - 70%) for
normal session lengths cannot tell true adjacencies from false adjacencies.
• Even with unrealistically long sessions (4 hours), the best method gets 90% accuracy for directions but finds very few adjacencies.
![Page 26: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/26.jpg)
26
Idea…
• If we could:– Increase sample size (effectively) by using data from
multiple subjects– Focus on a method with strong adjacencies– Combine this with a method with strong orientations
• We may be able to do better (Ramsey, Hanson and Glymour, NeuroImage)– This is the strategy of the PC Lingam algorithm of
Hoyer and several of us, though there are other ways to pursue the same strategy.
![Page 27: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/27.jpg)
27
Reminder: If noises are non-Gaussian, we can learn more than a pattern.
Linear Models, Covariance Data,Pattern/CPDAG
Linear Models, Non-Gaussian Noises (LiNG),
Directed Graph
(1) (2)
![Page 28: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/28.jpg)
28
Are noises for FMRI models non-Gaussian?• Yes. This is controversial but shouldn’t be.– For word/pseudoword data of Xue and Poldrack (Task 3),
kurtosis ranges up to 39.3 for residuals.• There is a view in the literature that noises are distributed
(empirically) as Gamma—say, with shape 19 and scale 20.
![Page 29: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/29.jpg)
29
Are connection functions linear for FMRI data?
• You tell me:
• I’ve not done a thorough survey of studies.
![Page 30: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/30.jpg)
30
Coefficients?• One expects them to be positive.– Empirically, in linear models of fMRI data, there are very few
negative coefficients (1 in 200, say).– They’re only slightly negative if so.– This is consistent with negative coefficient occurring due to
small sample regression estimation errors.• For the most part, need to be less than 1.– Brain activations are cyclic and evolving over time.– Empirically, in linear models of fMRI, most coefficients are
less than 1. To the extent that they’re greater than 1, one suspects nonlinearity.
![Page 31: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/31.jpg)
31
The IMaGES algorithm• Adaptation for multiple subjects of GES, a Bayes net
method tested by Smith, et al.• Iterative model construction using Bayesian scores
separately on each subject at each step; edge with best average score added.
• Tolerates ROIs missing in various subjects.• Seeks feed forward structure only.• Finds adjacencies between variables with latent
common causes.• Forces sparsity by penalized BIC score to avoid
triangulated variables (see Measurement Complexity)
![Page 32: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/32.jpg)
32
IMaGES/LOFS
• Smith (2011): “Future work might look to optimize the use of higher-order statistics specifically for the scenario of estimating directionality from fMRI data.”
• LiNGAM orients edges by non-Normality of higher moments of the distributions of adjacent variables.
• LOFS uses the IMaGES adjacencies, and the LiNGAM idea for directing edges (with a different score for non-Normality, and without independent components).
• Unlike IMaGES, LOFS can find cycles.• LOFS (from our paper) is R1 and/or R2…
![Page 33: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/33.jpg)
33
Procedure R1(S)
You don’t have to read these—I’ll describe them!• G <- empty graph over variables of S• For each variable V– Find the combination C of adj(V, S) that maximizes
NG(eV|C).– For each W in C• Add W V to G
• Return G
![Page 34: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/34.jpg)
34
Procedure R2(S)• G <- empty graph over variables of S• For each pair of variables X, Y
– Scores<-empty– For each combination of adjacents C for X and Y
• If NG(eX|Y) < NG(X) & NG(eY|X) > NG(Y)– score <- NG(X) + NG(eY|X)– Add <X->Y, score> to Scores
• If NG(eX|Y) > NG(X) & NG(eY|X) < NG(Y)– score <- NG(eX|Y) + NG(Y)– Add <X<-Y, score> to Scores
– If Scores is empty• Add X—Y to G.
– Else• Add to G the edge in Scores with the highest score.
• Return G
![Page 35: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/35.jpg)
35
Non-Gaussianity Scores
• Log Cosh – used in ICA • Exp = -e^(-X^2/2) – used in ICA• Kurtosis – ICA (one of the first tried, not great)• Mean absolute – PC LiNGAM• E(e^X) – Cumulant arithmetic
= e^(κ1(X) + 1/(2!) κ2(X) + 1/(3!) κ3(X) + …)
• Anderson Darling A^2 – LOFS – Empirical Distribution Function (EDF) score with heavy
weighting on the tails.– We’re using this one!
![Page 36: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/36.jpg)
36
Mixing Residuals
• We are assuming that residuals for ROIs from different subjects are drawn from the same population, so that they can be mixed.
• Sometimes we center residuals from different subjects before mixing, sometimes not.
• For Smith study, doesn’t matter—the data is already centered!
![Page 37: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/37.jpg)
37
Precision and Recall
• Precision = True positives / all positives– What fraction of the guys you found were correct?
• Recall = True positives / all true guys– What fraction of the correct guys did you find?
![Page 38: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/38.jpg)
38
![Page 39: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/39.jpg)
39
![Page 40: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/40.jpg)
40
![Page 41: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/41.jpg)
41
![Page 42: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/42.jpg)
42
![Page 43: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/43.jpg)
43
Some Further Problems
• Discovering nearly canceling 2 cycles is hard (but we will try anyway…)
• Identifying latent latents for acyclic models• Reliability of search may be worse with event
designs than with block designs• Subjects that differ in causal structures will
yield poor results for multi-subject methods.
![Page 44: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/44.jpg)
44
Hands On
• Download task3extracted.tet and task3extracted.zip. Make sure the zip file extracted if not automatically.
• Open the session.• Place a new Data box on the workbench.• Select Load Data from the File menu.• Using the shift key, select all of the files in the zip.• Load them.
![Page 45: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/45.jpg)
45
Hands On
• Attach a Search box to the data and run IMaGES.
• Copy the layout from the layout graph provided into the search box (using menus).
• Attach another Search box with IMaGES and Data as input and run LOFS.
![Page 46: Causal Modeling of FMRI Data](https://reader035.vdocuments.site/reader035/viewer/2022062410/56816207550346895dd230a2/html5/thumbnails/46.jpg)
46
Thanks!• S. M. Smith, K L. Miller, G. Salimi-Khorshidi, M. Webster, C. F.
Beckmann, T. E. Nichols, J. D. Ramsey, M. W. Woolrich (2011), Network modelling methods for fMRI, NeuroImage.
• J.D. Ramsey, S.J. Hanson, C. Hanson, Y.O. Halchenko, R.A. Poldrack, and C. Glymour(2010), Six Problems for causal inference from fMRI, NeuroImage.
• J.D. Ramsey, S.J. Hanson, C. Glymour. Multi-subject search correctly identifies causal connections and most causal directions in the DCM models of the Smith et al. simulation study. NeuroImage.
• G. Xue, Poldrack, R., 2007. The neural substrates of visual perceptual learning of words: implications for the visual word form area hypothesis. J. Cogn. Neurosci.
• Thanks to the James S. McDonnell Foundation.